We are using a global gene expression profiling approach based on state-of-the-art microarray technology to profile clinical specimens that are associated with different stages of liver diseases. For example, by comparing liver samples from chronic liver disease patients with varying degrees of risk for developing hepatocellular carcinoma, we have identified a unique signature that may be useful in diagnosing patients with early onset of liver cancer. Several serum proteins have been identified as potential diagnostic markers for hepatocellular carcinoma that are present at an early stage or in those negative for alpha-fetoprotein. We have also developed a unique molecular signature based on the mRNA gene expression of metastatic primary hepatocellular carcinoma specimens to predict prognosis and metastasis of hepatocellular carcinoma patients. We have recently validated the predictive capacity of this signature in two independent cohorts of differing etiology. Importantly, we found that this molecular signature could identify those patients who were most at risk for recurrence even in patients with early stage disease. Since hepatocellular carcinoma is usually present in inflamed liver, due to fibrosis, cirrhosis and/or chronic hepatitis, we also developed a unique molecular prognostic signature based on mRNA gene expression of the liver microenvironment of hepatocellular carcinoma patients. We found that a predominant humoral cytokine profile occurs in the metastatic liver microenvironment and that a shift toward anti-inflammatory/immune-suppressive responses may promote hepatocellular carcinoma metastases. These studies therefore suggest an important role of the local tissue microenvironment in hepatocellular carcinoma metastasis. Interestingly, the tumor signature is principally different from that of liver microenvironment. In addition, we have used molecular profiling to identify five genes that may serve as biomarkers for early onset of hepatocellular carcinoma, especially for those that are negative for alpha-fetoprotein. We have also explored the role of small non-coding RNAs, termed microRNAs, in hepatocellular carcinoma metastasis and survival. We found that certain microRNA expression changes are associated with metastasis and could significantly predict patient survival and relapse even in early stage disease. These microRNAs may provide a simple profiling method to assist in identifying HCC patients who are likely to develop metastases. In addition, functional analysis of these microRNAs may enhance our biological understanding of hepatocellular carcinoma metastasis. To further assess the role of microRNAs in the liver, we examined the microRNA expression patterns, survival and response to interferon in men and women with hepatocellular carcinoma. We found that the expression of microRNA-26 differed among men and women and was higher in tumor versus nontumor tissue. Tumors with reduced microRNA-26 expression had a distinct transcriptomic pattern with activation of the NFkB/IL6 signalling pathways. Patients with low microRNA-26 had poor survival and were better responders to interferon therapy than those with normal expression. We have recently followed up on this study by developing a multiplex reverse-transcription quantitative polymerase chain reaction(qRT-PCR) to determine the levels of two HCC-related microRNA-26 transcripts along with six small RNA reference transcripts. We evaluated archive paraffin-embedded tissues from three HCC cohorts (n=248) who underwent radical resection which was stratified into three groups (training and two test cohorts). We developed a matrix template and a scoring algorithm to assign patients into either low or high microRNA-26 groups. Patients with low microRNA-26 levels selected by the template were those that responded favorably to interferon-alpha therapy. We have demonstrated that this microRNA-26 diagnostic platform (MIR26-DX) is a simple and reliable diagnostictest to select patients for adjuvant interferon-alpha therapy. We have now initated a multi-center randomized control clinical trial in China based on these findings (NCT01681446). We have recently used an integrative approach to identify HCC driver genes, defined as genes whose copy numbers are associated with gene expression and cancer progression. Through a combination of data from high-resolution, arraybased comparative genomic hybridization and transcriptome analysis of HCC samples from 76 patients with hepatitis B virus infection, we found a 10-gene signature associated with chromosome 8p loss and poor outcome. The signature was validated in 2 independent HCC cohorts and breast cancer cohorts. Functional in vitro and in vivo studies demonstrated that three gene products among the 10-gene signature have tumor suppressive properties. Thus our integrative approach has identified driver genes that may assist in HCC diagnosis, prognosis and the development of new therapeutic strategies to improve HCC patient survival. Metabolites are the best molecular indicators of cell status because their rapid fluxes are an extremely sensitive measure of cellular phenotype. In this study, we sought to combine global metabolite and mRNA profiles to define key signaling events that can alter the fitness of EpCAM+ AFP+ HCC cancer stem cells. We analyzed frozen tissues of this tumor subgroup using a Metabolon platform, and found that this a sensitive method to discriminate tumor from paired non-tumor tissues. Furthermore, we identified 28 statistically significant metabolites that are associated with a stem cell-like HCC subgroup linked to patient survival. To determine how genes and biochemical molecules functionally interconnect, we performed a parallel analysis within the same tumor cohort to identify significant genes in the stem cell-like tumor subgroup. Following a correlativeness test of the metabolite-gene pairs, we performed an interdependency test by hierarchical clustering based on the correlation coefficient of each metabolite-gene pair to determine the connectivity between the principle metabolite-gene pairs. Interestingly, we found a metabolite cluster of fatty acids, suggesting a functional link. We then globally searched for surrogate genes for the fatty acid metabolites. Similar to the fatty acid metabolites, their 273 gene surrogates can predict HCC survival using an independent patient cohort. Finally, we performed gene network analysis and found that 273 genes are mainly associated with AKT-PI3K signaling. Interestingly, among this signaling network, we noticed SCD, which encodes stearoyl CoA desaturase, a key enzyme involved in fatty acid biosynthesis. SCD, which we found to be highly elevated in stem cell-like HCC, is responsible for converting the saturated FA, palmitate (aka SPA) to its monounsaturated form, palmitoleate (aka MUPA). We found that similar to SCD, both MUPA and methyl-SPA are highly elevated in stem cell-like HCC and are associated with HCC survival. Thus, it is possible that SCD and MUPA may functionally contribute to HCC stemness and aggressiveness. Consistently, we found that MUPA can stimulate while SPA can inhibit cellular invasion in vitro. Moreover, silencing of SCD can reduce HCC tumorigenicity in xenograft models. Our findings have been extremely fruitful and offer useful tools for personalized patient management and also challenge the current paradigm of tumor evolution. Clearly, expression profiling has expanded our knowledge of the global changes that occur in liver cancer, and has provided numerous insights into the molecular mechanisms of this disease. In addition, these studies will undoubtedly contribute to the establishment of novel markers with potential diagnostic and prognostic value, as well as potential therapeutic targets for direct clinical intervention.